Remember those long holiday road trips where the question endlessly repeated was “are we there yet”? Well, for many in the peak oil community, waiting for it to arrive has evoked a similar feeling, as the predictions of some academics, commentators, and bloggers have failed to materialize punctually. So, it’s worth revisiting some thoughtful papers that reveal a more sophisticated approach to timing the peak date. This may give people a greater appreciation of the challenges of predicting such a complex event.

Various mathematical models have been evoked to relate production to estimations of the total amount of oil that can be extracted, also known as the ultimately recoverable resource (URR). These include linear, exponential and, most famously, Hubbert’s logistical model. Attempts have been made to determine the best fit of these models to historical production. However, in terms of prediction, none are necessarily more correct than the others. The outcomes they predict are strongly determined by the assumptions and parameters used to shape the resultant curves.

No doubt, the size of the assumed URR can have a large effect on the date of the peak. But equally so can the values for the rate of oil production growth and decline which determine the symmetry of the production curve. A higher rate on the decline side creates a production curve skewed to the right, and vice versa.

It was with this in mind that researchers at the US Energy Information Administration (EIA) attempted to use mathematical modelling to predict a date for a peak in global oil production1 (see Fig. 1). For the sake of simplicity we’ll consider their analysis, and some others, with a mid-range estimation of URR of 3000 Gb. For a realistic growth rate of 2% annually they made a prediction of a peak in 2037. In doing this they made an assumption of a reserves to production ratio of 10 (i.e. decline rate of 10% annually), which was based on US historical production experience.

However, the decline rate assumptions by the EIA researchers left them open to criticism by later modellers as being unacceptably high. This was because the US decline rate they used was based on proven reserves and not on the larger estimations for URR for the US. These researchers claimed that the decline rate in reality should have been much lower. This would have, in effect, shifted the estimated peak prediction years earlier.

More recent analyses have instead been based on various growth and decline rates, leading to a range of prediction dates for a peak in production. For simplicity, and to enable a comparison among these and the earlier EIA model, we’ll keep our assumption of a URR of approximately 3000 Gb and a growth rate of around 2%. This growth rate is close to that observed globally so far. When we make these less restrictive assumptions and use decline rates realistically varying from 2 – 6%, we get a time range of 2019 – 20262 and 2013 – 20333 from the two studies, respectively (see Fig. 2).

These results cast a different light, not only on the prediction of a peak in oil production, but indeed on our ability to precisely predict the date. It is worth noting that many early peak oil predictions have been based on symmetrical production curves where production growth and decline rates were equal. This is a reasonable assumption for prediction when the actual future values for these rates are unknown. However, in reality, production curves are rarely symmetrical.

Related to these comments, an analysis by HSBC Bank in 2016 indicated that, due to the nature of oil production where larger fields are exploited first, later decline rates may be higher than those used in symmetrical Hubbert modelling. Their conclusion was that they may in fact be 6% on average. So, whereas a prediction of 2037 based on a 10% decline rate for mid-range URR may be improbably late, a prediction before 2020 based on relatively low decline rates may be improbably early. One of the above studies showed that of 64 global production scenarios 53 gave a peak before 20302, so we can be reasonably confident of a peak before then.

Admittedly, the above interpretation of these studies is a simplification, but it does illustrate a couple of points. Firstly, for any given assumption of URR, there can be a large range of predicted global production dates dependent on the assumptions of growth and decline rates that were made. Hence, we should not expect a punctual prediction of a peak date due to the large errors inherent not only on our limited knowledge of URR, but also production, which may affect the real date by many years.

Additionally, it also explains the large gap in a date made for peak oil by different researchers, public energy agencies and informed commentators. So our impatient holidaymakers may have to patiently wait a little longer for the reality of peak oil in a finite world!


  1. Wood J.H., Long G.R. and Morehouse D.F. (2004). Long-term world oil supply scenarios: the future is neither as bleak or rosy as some assert. Energy Information Administration.
  2. Kaufmann R.K. and Shiers L.D. Alternatives to conventional crude oil: When, how quickly, and market driven? Ecological Economics 67: 405 – 411.
  3. Sorrell S., Miller R., Bentley R. and Speirs J. (2010). Oil futures: A comparison of global supply forecasts. Energy Policy 38: 4990–5003.